群吧 plot 合 Pandas plot

Question

I am new with Python plotting, and I am trying to group bars together using a pandas dataframe.

my input csv looks like this-- (test.csv)

noc,matrix,perf
jump4,dw,7.5
jump4,t2d,7.7
mesh,dw,2
mesh,t2d,4

and python plotting script is--

import matplotlib.pyplot as plt
import pandas as pd
from matplotlib.backends.backend_pdf import PdfPages
import numpy as np

df = pd.read_csv("test.csv", sep=",")
pp = PdfPages('rtl_sim_perf.pdf')

# gca stands for 'get current axis'
rc = {"axes.spines.top"    : False,
      "axes.spines.right"  : False,
      "axes.edgecolor"     : "black"}
plt.style.use(("ggplot", rc))
ax = plt.gca()

df.plot(x='matrix', y=['perf','noc'], ax=ax,kind='bar', edgecolor='k')
plt.xticks(fontsize=5,rotation=0)
plt.yticks(fontsize=5)
plt.legend(fontsize=7)
plt.tight_layout()
plt.savefig(pp, format='pdf')
pp.close()

and I get this result

What I want is to group bars based on matrix column, so that dw are together for different noc and t2d together for different noc . And the legend shows me noc type. I am not sure how to do this.

Answer 1

IIUC, do you want?

df.set_index(['matrix','noc'])['perf'].unstack().plot.bar()

Output:

You can achieve this by reshaping your dataframe such that you have 'noc' as columns in your dataframe and using pandas plot.

---

Another way is using seaborn and you don't have to reshape the dataframe and use the 'hue' parameter.

import seaborn as sns
sns.barplot(x='matrix', y='perf', hue='noc', data=df)

Output: